Source code for cpsim.attack

import numpy as np




class Attack:
    def  __init__(self, category, param, start_index, end_index=None):
        """
        category: 'bias', 'delay', 'replay'  todo: 'stealthy'
        param
        """
        self.cat = category
        self.param = param
        self.start_index = start_index
        self.end_index = end_index if end_index is not None else np.Inf
        self.cur_index = None
        self.disabled = False



    def update_param(self, param):
        self.param = param




    def disable(self):
        self.disabled = True




    def launch(self, cur_data, cur_index, history_intact_data):
        self.cur_index = cur_index
        if cur_index < self.start_index or cur_index > self.end_index or self.disabled:
            return cur_data
        if self.cat == 'bias':
            # param is the bias on each element
            return cur_data + self.param
        if self.cat == 'delay':
            # param is the number of delay steps
            if self.cur_index - self.start_index > self.param:
                return history_intact_data[self.cur_index - self.param]
            else:
                return history_intact_data[self.start_index]
        if self.cat == 'replay':
            # replay interval param['start'] ~ param['end']
            delta_index = self.cur_index - self.start_index
            offset_index = self.param['start'] + delta_index % (self.param['end']-self.param['start'])
            return history_intact_data[offset_index]




    def get_C_filter(self):
        if self.cat == 'bias':
            cnt = len(self.param)
            bias_e = np.full((cnt,), False)
            for i in range(cnt):
                if self.param[i] != 0:
                    bias_e[i] = True
            C_filter = []
            for i in range(4):
                if bias_e[i]:
                    continue
                else:
                    tmp = [0, 0, 0, 0]
                    tmp[i] = 1
                    C_filter.append(tmp)
            return C_filter
        else:
            raise NotImplementedError()




if __name__ == "__main__":
    attack = Attack('bias', np.array([1, 0, 0, 0]), 0)
    print(attack.get_C_filter())